3 ms·
> it's probably not 1/273, it's probably 1/n where n > 273. This is equally puzzling to me. It sounds like you're saying outliers are more likely to be extreme
by morelisp 4y ago
> it's probably not 1/273, it's probably 1/n where n > 273.
This is equally puzzling to me. It sounds like you're saying outliers are more likely to be extreme outliers than less extreme outliers, which is... not true.
- samatman 4y agoHopefully this will help: The real number could plausibly be 1/250, after 10,000 samples. More than that would really call for an explanation: if it was 1/100, why didn't we find another one in the first 272? Also, the real number could plausibly be 1/100,000.0, we've never seen a counterexample. There's nothing bounding the lower end of the range. The distribution of these probabilities is substantially lower than 273. You can work it out precisely with Bayes' Theorem, if you'd like, this is a verbal explanation of how that sort of statistical reasoning proceeds.
- morelisp 4y ago> The distribution of these probabilities is substantially lower than 273. There's no evidence for this, you can't just point to "there's a lot of numbers and we only saw 273 of them" to justify it. The distribution seems exactly what we're trying to determine; to me it sounds like a good example of when Bayesianism breaks down.